magnetic resonance imaging
Springer Online Journal Archives 1860-2000
Abstract Recently the use of stereotactic radiosurgery to treat functional disorders such as Parkinson's disease, epilepsy, and intractable pain has been reported in the literature. In such applications, a large single dose is typically delivered to an extremely small (〈0.05 cm3) target volume. The purpose of this work is to investigate whether the dosimetric and imaging characteristics of radiosurgery treatment planning provide sufficient accuracy to allow efficacious therapy of functional disorders. We have begun treating trigeminal neuralgia using our linear accelerator-based radiosurgery system: 70 Gy is prescribed to the maximum dose in the volume (in our case the 100% isodose level) and delivered to the base of the fifth nerve in a single fraction using a 5 mm collimator, with the standard Brown–Roberts–Wells (BRW) radiosurgery accessories employed for fixation and localization. Because the fifth nerve cannot be visualized on x-ray computed tomography (CT), our radiosurgery treatment planning system was modified to use magnetic resonance images for localization, though dose calculations are still performed using CT. Isocentric accuracy of our original radiosurgery system, consisting of a floor stand and isocentric subunit, and our new couch mount system, was evaluated using the Winston–Lutz film test method. In order to evaluate the spatial accuracy of magnetic resonance (MR) treatment planning, eight 4 mm sections of a 7 French catheter were filled with CuSO4 contrast material and attached rigidly to the stereotactic fixation posts of our BRW frame, four each in orientations parallel and perpendicular to the axial plane. The position of the externally placed fiducial markers, as well as internal anatomical structures, were then compared with CT. Monte Carlo calculations were compared with those from a commercial radiosurgery treatment planning system in order to investigate the effects of tissue heterogeneities on the resulting dose distributions. While commercial radiosurgery systems assume tissue homogeneity, the Monte Carlo calculations were performed in a patient-specific CT geometry accounting for all tissue inhomogeneities. The resulting 128 × 128 Monte Carlo dose grid was superimposed on the original CT data for analysis and comparison with identical treatment plans from the commercial system. The ability of our LINAC-based systems to accurately target a desired point in space has been effectively demonstrated: 0.32 ± 0.32 mm (N = 556) for our floor stand system and 0.34 ± 0.23 mm (N = 50) for our newer couch-mounted system. Inaccuracies introduced by tomographic imaging devices are significantly greater. The use of gel-filled fiducial markers in magnetic resonance imaging (MRI) guided radiosurgery produces significant spatial distortion, resulting in Euclidean root-mean-square deviations of 2.32 ± 0.96 mm (N = 31) and 3.64 ± 1.28 mm (N = 27) at the center and periphery (extracranial) of the field of view respectively, as compared with CT. Use of water of CuSO4 filled rods had a minimal effect on these deviations: 2.51 ± 1.25 mm (N = 31) and 3.37 ± 1.28 mm (N =27) for central and peripheral targets respectively. Magnetic susceptibility artifacts in the frequency encoding (AP) direction produce a systematic posterior shift. This together with axial slice spacing accounts for the majority of the deviation. Tissue heterogeneities such as bone and air cavities produce a lateral spreading of the dose from small photon beams, resulting in a prescription dose volume smaller than predicted by conventional treatment planning systems. For a typical are configuration designed to produce a spherical dose volume, Monte Carlo calculations show the 90% dose volume to be significantly smaller than that predicted by the commercial system when either 5 mm or 10 mm collimators are used. Use of a LINAC-based system does not preclude accurate treatment of small functional targets. Isocentric uncertainty for either of two LINAC systems that we evaluated is small compared to imaging and dosimetric factors. However, chemical shifts and object-induced magnetic susceptibility artifacts can produce systematic spatial distortions in magnetic resonance images; thus, MR imaging may not possess the inherent accuracy necessary for stereotactic localization and targeting of small cranial structures. In addition, both CT and MR possess an inherent inaccuracy of at least one-half of the axial slice thickness; thus, for localization purposes, a slice spacing as small as possible should be used when treating small targets. Tissue heterogeneities decrease the volume covered by the higher isodose lines. As a result, the target may be only partially covered by the intended dose level, with the remainder lying in the high gradient region. This same lateral spreading may also increase the risk to adjacent normal structures. Imaging and dosimetric considerations are not unique to linear accelerator systems but apply equally to all stereotactic photon irradiation. Until spatial and dosimetric errors can be accounted for, use of a larger collimator will ensure better coverage of small targets.
Type of Medium: